| Many animals, e.g., lobsters, blue crabs, ants and moths, are capable of tracking chemical plumes to their source. This ability is not trivial, because the turbulence of the fluid medium (water or air) continuously stretches and twists the filaments of the chemical plume and changing flow vectors causes the plume centerline to meander. Autonomous Underwater Vehicles (AUVs) capable of such chemical plume tracing feats would have applicability in searching for environmentally interesting phenomena, unexploded ordnance, undersea wreckage, and sources of hazardous chemicals or pollutants.; This thesis considers the development and field test of reactive planning algorithms to replicate the animals' chemical plume tracing ability using a REMUS AUV. The goal of the AUV will be to locate sources of chemical in a turbulent fluid flow. In order to plan behaviors to effectively complete the chemical plume tracing mission, AUVs need some representation of the environment, in general a map. The problem of autonomous vehicle mapping has received considerable attention over the past few years. This thesis develops new online mapping algorithms based on Hidden Markov Methods (HMM's). Given an estimate of the source location, HMM's provide efficient algorithms for predicting likelihood of odor detection versus position, the most likely path taken by the odor to a given location, and the path between two points that is most likely to result in odor detection. All three are useful for solving the chemical plume tracing problem. A new computational efficient algorithm to estimate likely chemical source locations was also developed using stochastic process and probability theory. This methodology results in algorithms for predicting likely source locations based on the measured flow and the detection or non-detection events that happened at the vehicle locations.; The above theoretical methods were too computationally intensive for on-vehicle implementation. Therefore, the new reactive mission planning approach was designed based on the behavior-based planning method. Six behaviors were designed (goto, find, track-in, track-out, reacquire, and cage) for plume tracing. A guidance system was also developed for REMUS AUV. Three sets of in-water field tests demonstrated the AUV plume tracing ability over distances of 975 m with average source declaration accuracy of approximately 13 m. |
